Pith. sign in

REVIEW 3 major objections 3 minor 51 references

Machine learning prediction of binary formation in three-body gravitational encounters

T0 review · 3 major / 3 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that binary formation in three-body encounters can be predicted from initial conditions alone by a gradient-boosted classifier, with all metrics above 0.94 and a speedup of about 400 over direct integration.

desk verdict A credible proof-of-concept for ML classification of binary formation in three-body encounters, but the headline accuracy applies only to a curated balanced test set; the paper's own Table 3 highlights a much weaker raw-data performance that the abstract glosses over. read the letter →

arxiv 2607.16776 v1 pith:Z6VK3WQG submitted 2026-07-18 astro-ph.GA astro-ph.SR

Ahmad Farhani Asl (1) ((1) Institute for Advanced Studies in Basic Sciences (IASBS) , Zanjan , Iran) This is my paper
classification astro-ph.GAastro-ph.SR
keywords three-bodyproblembinaryformationmachinelearninggradientboostingstellardynamicsN-bodyscatteringhardnessratiofeatureimportance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the discrete outcome of a chaotic three-body gravitational encounter—whether a binary forms—can be predicted accurately and almost instantly from the initial conditions alone, without integrating the orbits. If true, large-scale stellar-dynamics simulations could replace many expensive direct integrations with a surrogate model, and the model's feature rankings would provide a quantitative statement of which physical parameters actually govern binary formation. The evidence is a gradient-boosted classifier trained on about 313,000 cleaned scattering experiments, balanced across hardness, that achieves a PR-AUC near 0.99 on a held-out test set and an inference speedup of roughly 400 times. A sympathetic reader would care because binary formation rates feed directly into cluster evolution, population synthesis, and predictions of compact-object merger sources.

What carries the argument

The load-bearing object is a gradient-boosted decision-tree ensemble trained as a binary classifier on a balanced dataset of three-body scattering simulations. The input representation carries much of the argument: each encounter is summarized by 30 features built from initial masses, velocities, impact parameters, and composite dynamical quantities—most importantly the mass hierarchy (the ratio of the most massive star to the least massive star) and the hardness ratio H = |U_pot|/U_kin, which measures how deeply bound the encounter is relative to its kinetic energy. The labels come from direct N-body integrations with a high-order adaptive integrator; a 'formed binary' means a pair with neg

What would settle it

Take a fresh sample of encounters and label binaries by a much longer integration (for example 100 times the encounter time) with no semimajor-axis or eccentricity cuts, then measure the classifier's PR-AUC and calibration on that relabelled set; a material drop from 0.99 would show the model is reproducing the filtering pipeline rather than the physical formation probability. A complementary probe is to generate encounters with hardness in [0.9,1), a regime absent from training, and compare predicted probabilities against direct integrations there.

Watch

Extended reading notes

Core claim

The central claim is that binary formation in an initially unbound three-body encounter is a learnable function of the initial conditions. A gradient-boosted tree ensemble on 30 physical features (mass hierarchy, hardness, velocities, impact parameters, focusing, angular momentum) reproduces direct-integration binary labels. On a balanced test set: accuracy/precision/recall/F1 > 0.94, ROC-AUC and PR-AUC near 0.99, calibration error 0.02, inference ~400x faster than direct integration. Feature importance puts mass hierarchy first and hardness second, with geometry secondary. Performance drops only for hardness < 0.01, where binary formation is rare and class imbalance severe.

Load-bearing premise

The load-bearing premise is that the paper's operational definition of a formed binary—a pair with negative pairwise energy that survives to ten times the encounter time and passes cuts on energy error, semimajor axis, eccentricity, and completed orbital periods—is a faithful proxy for astrophysically real binaries; if those filters exclude genuine wide or eccentric binaries, the model has learned to predict a narrower, partially artefactual target.

Editorial extensions

If this is right

  • Large-scale simulations of dense stellar systems can use the classifier as a fast surrogate for binary-formation probability, replacing many direct integrations at a speedup of roughly 400 times.
  • The feature-importance hierarchy gives a quantitative answer to an old dynamical question: mass hierarchy and hardness dominate the outcome, while encounter geometry is secondary.
  • Because the predicted probabilities are well calibrated (ECE 0.02), the model can feed Monte Carlo sampling in cluster evolution and binary population synthesis.
  • The model generalizes across encounter radii spanning open clusters to dense cores, with a known limitation in the weak-interaction regime where binary formation is intrinsically rare.
  • The gap over single-feature baselines (PR-AUC 0.99 versus 0.72) shows that binary formation depends on nonlinear combinations of features, not simple thresholds.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same feature set could be extended to predict the properties of the formed binary—semimajor axis, eccentricity, and mass ratio—since the dynamical invariants that set those distributions are already partly encoded in the initial conditions; the paper's own orbital statistics suggest such a regression target is well-populated.
  • The dominance of mass hierarchy may reflect phase-space volume rather than new dynamics: the classifier could be learning which mass combinations have the largest pairing phase space, a hypothesis testable by comparing its decision boundary to analytic probability estimates in the near-unbound limit.
  • Applying the model to realistic stellar populations will require reweighting the log-uniform training distribution; a concrete test is to retrain on a mass function and velocity dispersion from a specific cluster model and compare predicted formation rates with direct integrations on that same population.
  • The degraded low-hardness performance could be probed by hardness-conditioned undersampling; if PR-AUC recovers to 0.99, the failure is class imbalance, whereas if it stays low, the feature set genuinely loses separability for weakly bound encounters.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. The paper trains an XGBoost classifier to predict whether an initially unbound three-body gravitational encounter forms a binary. A dataset of about 313,000 REBOUND/IAS15 scattering simulations is generated with stratified sampling in hardness; 30 physically motivated features are constructed from the initial conditions. The model is trained on a balanced (undersampled) dataset and evaluated on a held-out balanced test set, reporting accuracy, precision, recall, and F1 above 0.94, ROC-AUC and PR-AUC of 0.99, ECE 0.02, and a roughly 400x speedup over IAS15. Feature importance ranks mass hierarchy and hardness first. The paper also reports performance on a newly generated imbalanced dataset and discusses degradation in the weak-interaction regime.

Significance. If the central claims hold, this is a useful contribution: a fast, interpretable surrogate for binary-formation outcome prediction, with potential application to Monte Carlo cluster simulations and population synthesis. The paper is clearly structured, uses a physically motivated feature set, and reports a comprehensive set of metrics. However, the manuscript cannot currently be fully assessed because code and data are not provided (the GitHub link is incomplete), and the reporting in Table 3 raises serious ambiguities that must be resolved. The central 'excellent performance' claim also needs to be qualified by the balanced-test-set context and by the much lower performance on the realistic imbalanced dataset (PR-AUC 0.66).

major comments (3)
  1. [Table 3 vs §3.1/Fig. 2] The central performance claim rests on the balanced test-set metrics in §3.1 and Fig. 2 (accuracy 0.944, recall 0.940, PR-AUC 0.99). Table 3 reports a 'Balanced dataset' with accuracy 0.82, recall 0.66, and PR-AUC 0.96 (88,750 samples). If this is the same balanced test set, the numbers are irreconcilable. The likely resolution is that Table 3 refers to the new randomly generated dataset from §3.4 (827,000 samples, binary fraction 0.05) and its balanced subset, not the stratified balanced test set. Please re-label Table 3 and state explicitly that the random dataset was used only for generalization/speed benchmarking, and that the >0.94 metrics apply only to the original balanced test set. If Table 3 is instead meant to report the main evaluation, the headline claim must be revised downward.
  2. [§2.2.2] The target class is defined by four quality cuts after simulation: |ΔE/E|<1e-8, a<1e5 au, e<0.9999, and at least two orbital periods. These exclude a substantial fraction of binary-forming events (4.7% by semi-major axis, 13.8% by eccentricity, 25% by the orbital-period criterion, 3.5% transient). The model is therefore trained to predict a filtered, astrophysically curated class, not raw binary formation. The paper should quantify the overlap among these cuts, state how the excluded events are distributed in parameter space (e.g., wide, eccentric, late-forming binaries), and justify that the filtered class is the physically meaningful target. If many genuine binaries are discarded, the reported metrics overstate performance on raw encounters.
  3. [§6 (Data Availability)] The manuscript states that code is 'publicly available on GitHub at /gtb' — the URL is incomplete. Data are to be 'deposited at the CDS upon acceptance.' Given that the reported metrics and Table 3 cannot be verified without code and data, a working repository link and a clear data-availability statement are essential. Please provide a complete URL and, if possible, deposit the data and training/evaluation scripts with the submission so the reported numbers can be checked.
minor comments (3)
  1. [§4] The sentence 'The identical recall and ROC-AUC across raw and balanced datasets' is inaccurate: Table 3 lists recall 0.65 and 0.66. Change to 'nearly identical' or update the table.
  2. [§3.4] The text says the new random dataset was 'neither used it for training nor for evaluation,' but Table 3 reports performance on it. Rephrase to clarify that this dataset was not used for training or for the main test-set evaluation, but was used for a separate generalization/speed benchmark.
  3. [Abstract and §1] The abstract and introduction state that the classifier 'achieves excellent performance' without specifying that this refers to the balanced test set with stratified sampling. Add a qualification such as 'on a balanced test set' and note the lower PR-AUC on the realistic imbalanced distribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: target labels come from independent N-body integrations and the classifier is evaluated on held-out data.

full rationale

The paper's claimed prediction—binary formation from initial conditions—is generated by a supervised classifier whose target labels are defined in §2.1.2 independently of the feature set: a binary is identified as the pair with lowest pairwise energy E_ij at t_max, verified by extending integration to 10 t_max and applying the four quality cuts in §2.2.2. The 30 features in §2.2.1 are computed from initial masses, velocities, impact parameters, angles, and composite quantities; none is fitted to the labels. Model performance is measured on a held-out 20% test set plus 10-fold cross-validation, so the reported accuracy/PR-AUC is an out-of-sample statement, not a restatement of training data. Feature importance and calibration are descriptive post-hoc analyses. The virial-radius cutoff feature is 'inspired by' Ginat & Perets (2024), an external analytic result, not a self-citation; no uniqueness theorem or prior self-result is invoked to force the model choice. The internal inconsistency between §3.1/Fig. 2 and Table 3 is a reporting/correctness problem, not circularity, since both surfaces claim to describe the same held-out data. The only definitional ambiguity is Table 2's ℓ_bin = L/|U_bin|, which is described in language about the 'final binary'; however, U_bin is never defined, and the paper states all features derive from initial conditions. Treating it as leakage would require speculation, which the circularity standard excludes. Therefore no circular step is established; score 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central result rests on simulation fidelity, the operational binary label, the balanced training distribution, and the hand-crafted feature set. No new physical entities are introduced. The main burden is that the model is trained and evaluated on a filtered, balanced dataset; the raw-data evaluation shows substantially lower PR-AUC.

free parameters (4)
  • XGBoost hyperparameters = n_estimators=1400, max_depth=18, learning_rate=0.02, subsample=0.9, colsample_bytree=1.0, lambda=1.0
    Selected by random search and manual tuning (§2.2); they control model capacity and regularization and were optimized on validation data, so the reported performance is tuning-dependent.
  • Quality-cut thresholds = |ΔE/E|<1e-8; a<10^5 au; e<0.9999; ≥2 orbital periods; extension to 10 t_max
    Post-hoc filters in §2.2.2 that define the cleaned dataset and the binary-formation target. The choices are physically motivated but not derived from first principles, and they materially change the class balance and separability.
  • Class balancing ratio = 1:1 positive:negative
    Random undersampling of the majority class (§2.2.2) was chosen because it outperformed class weighting on validation metrics. This balanced distribution is used for the headline metrics and inflates them relative to the natural 5% binary fraction.
  • Binary classification decision threshold = 0.5 (implied)
    Precision, recall, F1, and accuracy in §3.1 and Table 3 depend on the probability threshold, which is not explicitly described. The discrepancy between confusion-matrix metrics and Table 3 suggests different thresholds or datasets are being compared without explanation.
assumptions (5)
  • domain assumption Newtonian point-mass gravity with G=4π²; no post-Newtonian corrections or physical collisions
    Stated in §2.1; limits applicability to nonrelativistic, non-merging encounters and may produce different binary-pairing statistics than finite-size/PN simulations (Atallah et al. 2024).
  • domain assumption IAS15 with relative energy error <1e-8 produces correct encounter outcomes
    Invoked in §2.1 and §2.2.2; 4% of runs were removed for exceeding the energy-error threshold. The quality of all labels depends on the integrator correctly resolving chaotic close encounters.
  • ad hoc to paper A pair is a formed binary iff E_ij<0 at final time and the pair survives extended integration with at least two orbital periods
    Operational definition in §2.1.2 and §2.2.2. It excludes transient, very wide, and highly eccentric binaries; if this does not match astrophysical binary lifetimes, the model learns a proxy class.
  • domain assumption Stratified hardness sampling and random undersampling produce a distribution from which a classifier can generalize to natural encounter populations
    The model is trained and evaluated on a balanced distribution (§2.1.1, §2.2.2), but the raw new random dataset has a 5% binary fraction and PR-AUC drops to 0.66, showing this generalization assumption is only partially satisfied.
  • ad hoc to paper The 30 hand-crafted features capture all physically relevant initial-condition information
    Feature set defined in §2.2.1. There is no proof that relative phases, full angular momentum distribution, or other details are unnecessary; the future-work section acknowledges additional features could help.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine learning prediction of binary formation in three-body gravitational encounters." pith.science (2026). https://pith.science/paper/Z6VK3WQG

@misc{pith2026260716776,
  author       = {Pith},
  title        = {Pith review of: Machine learning prediction of binary formation in three-body gravitational encounters},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z6VK3WQG}},
  note         = {Machine review of arXiv:2607.16776}
}
read the original abstract

Three-body encounters are frequent events in stellar systems, intrinsically chaotic, and computationally costly to model with direct N-body integration. Predicting whether such encounters lead to binary formation is therefore challenging, particularly in large-scale simulations. We aim to develop an accurate and physically interpretable machine-learning model that predicts binary formation from the initial conditions of three-body encounters, to assess its reliability, and to identify the physical parameters that most strongly determine the outcome. We trained an XGBoost binary classifier on a balanced dataset of three-body scattering experiments computed with the REBOUND code and the IAS15 integrator. The input to the model consists of 30 physically motivated features describing the masses, energies, and kinematics of the initial configuration. The classifier achieves excellent performance on a balanced test set, with accuracy, precision, recall, and F1-score all above 0.94, and with ROC-AUC and PR-AUC values of 0.99. Feature-importance analysis shows that the outcome is governed primarily by the mass hierarchy and hardness ratio of the encounter, followed by velocity fraction and mass entropy. The predicted probabilities are well calibrated, with an expected calibration error (ECE) of 0.02. Inference is approximately 400 times faster than direct N-body integration. The model also generalizes well across encounter radii, although its performance decreases in the weak-interaction regime where binary formation becomes intrinsically rare. These results show that machine learning can provide fast, accurate, and physically interpretable predictions of binary formation in three-body encounters. Such models offer a practical complement to direct N-body simulations and may enable efficient probability estimates in large-scale simulations of stellar systems.

Figures

Figures reproduced from arXiv: 2607.16776 by the authors.

Figure 1
Figure 1. Initial configuration for a three-body encounter. Each star [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Confusion matrix evaluated on the test set. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Top important features of the XGBoost classifier. 1 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 5 linked inside Pith

  1. [1]

    Aarseth, S. J. 1999, PASP, 111, 1333

    1999

  2. [2]

    2025, Master’s thesis, Stellenbosch University, Stellenbosch, South Africa, mEng Thesis

    Albertyn, C. 2025, Master’s thesis, Stellenbosch University, Stellenbosch, South Africa, mEng Thesis

    2025

  3. [3]

    2022, MNRAS, 516, 4716

    Alegre, L., Sabater, J., Best, P., et al. 2022, MNRAS, 516, 4716

    2022

  4. [4]

    B., & Weatherford, N

    Atallah, D., Ginat, Y . B., & Weatherford, N. C. 2025, ApJ, 993, 183

    2025

  5. [5]

    C., Trani, A

    Atallah, D., Weatherford, N. C., Trani, A. A., & Rasio, F. A. 2024, ApJ, 970, 112

    2024

  6. [6]

    & Bengio, Y

    Bergstra, J. & Bengio, Y . 2012, Journal of Machine Learning Research, 13, 281

    2012

  7. [7]

    & Tremaine, S

    Binney, J. & Tremaine, S. 2008, Galactic Dynamics: Second Edition (Princeton, NJ: Princeton University Press) Bojnordi Arbab, B. & Rahvar, S. 2021, International Journal of Modern Physics D, 30, 2150063

    2008

  8. [8]

    G., Foley, C

    Breen, P. G., Foley, C. N., Boekholt, T., & Portegies Zwart, S. 2020, MNRAS, 494, 2465

    2020

Show all 51 references
  1. [9]

    2020, arXiv e-prints, arXiv:2008.11107 Caritá, G

    Capuzzo-Dolcetta, R., De Cesare, G., & Marino, A. 2020, arXiv e-prints, arXiv:2008.11107 Caritá, G. A., Aljbaae, S., Morais, M. H. M., et al. 2024, Celestial Mechanics and Dynamical Astronomy, 136, 10

    2020

  2. [10]

    2004, Using Random Forest to Learn Imbal- anced Data, Tech

    Chen, C., Liaw, A., & Breiman, L. 2004, Using Random Forest to Learn Imbal- anced Data, Tech. Rep. 666, University of California, Berkeley 110 (1-12): 24

    2004

  3. [11]

    & Guestrin, C

    Chen, T. & Guestrin, C. 2016, arXiv e-prints, arXiv:1603.02754

    2016 arXiv

  4. [12]

    Davies, M. B. 1996, in IAU Symposium, V ol. 174, Dynamical Evolution of Star Clusters: Confrontation of Theory and Observations, ed. P. Hut & J. Makino, 243 De Cesare, G. & Capuzzo-Dolcetta, R. 2021, Ap&SS, 366, 53

    1996

  5. [13]

    Fregeau, J. M. & Rasio, F. A. 2007, ApJ, 658, 1047

    2007

  6. [14]

    2018, Journal of Intelligent Tech- nologies and Applied Statistics, 11, 105

    Gholamy, A., Kreinovich, V ., & Kosheleva, O. 2018, Journal of Intelligent Tech- nologies and Applied Statistics, 11, 105

    2018

  7. [15]

    Ginat, Y . B. & Perets, H. B. 2024, MNRAS, 531, 739

    2024

  8. [16]

    Hairer, E., Nørsett, S., & Wanner, G. 1993, Solving Ordinary Differential Equa- tions II: Stiffand Differential-Algebraic Problems, Solving Ordinary Differ- ential Equations II: Stiffand Differential-algebraic Problems (Springer)

    1993

  9. [17]

    M., Clarke, C

    Hall, S. M., Clarke, C. J., & Pringle, J. E. 1996, MNRAS, 278, 303

    1996

  10. [18]

    & Modak, S

    Hamilton, C. & Modak, S. 2024, MNRAS, 532, 2425

    2024

  11. [19]

    & Hut, P

    Heggie, D. & Hut, P. 2003, The Gravitational Million-Body Problem: A Multi- disciplinary Approach to Star Cluster Dynamics (Cambridge, UK: Cambridge University Press)

    2003

  12. [20]

    Heggie, D. C. 1975, MNRAS, 173, 729

    1975

  13. [21]

    C., Hut, P., & McMillan, S

    Heggie, D. C., Hut, P., & McMillan, S. L. W. 1996, in IAU Symposium, V ol. 174, Dynamical Evolution of Star Clusters: Confrontation of Theory and Ob- servations, ed. P. Hut & J. Makino, 371

    1996

  14. [22]

    1985, IAU Symposium, 113, 231, binary-single star scattering experi- ments

    Hut, P. 1985, IAU Symposium, 113, 231, binary-single star scattering experi- ments

    1985

  15. [23]

    & Bahcall, J

    Hut, P. & Bahcall, J. N. 1983, ApJ, 268, 319

    1983

  16. [24]

    M., & Rasio, F

    Ivanova, N., Belczynski, K., Fregeau, J. M., & Rasio, F. A. 2005, MNRAS, 358, 572 Jiménez-Torres, J. J., Pichardo, B., Lake, G., & Segura, A. 2013, Astrobiology, 13, 491

    2005

  17. [25]

    2001, MNRAS, 322, 231

    Kroupa, P. 2001, MNRAS, 322, 231

    2001

  18. [26]

    2020, MNRAS, 493, 1842

    Lin, H., Li, X., & Luo, Z. 2020, MNRAS, 493, 1842

    2020

  19. [27]

    Lorenz, E. N. 1963, Journal of the Atmospheric Sciences, 20, 130 Machado Poletti Valle, L. F., Avestruz, C., Barnes, D. J., et al. 2021, MNRAS, 507, 1468

    1963

  20. [28]

    Makarov, V . V . 2025, AJ, 170, 138 Manoel Herrera Pinheiro, J., Vilas Boas de Oliveira, S., Segreto Silva, T. H., et al. 2025, arXiv e-prints, arXiv:2506.08274

    2025

  21. [29]

    & Kroupa, P

    Marks, M. & Kroupa, P. 2012, A&A, 543, A8

    2012

  22. [30]

    S., Xu, S., Federrath, C., Hu, Y ., & Seta, A

    Mathew, S. S., Xu, S., Federrath, C., Hu, Y ., & Seta, A. 2024, MNRAS, 532, 2374

    2024

  23. [31]

    2021, arXiv e-prints, arXiv:2111.03306

    Mojiri, A., Khalili, A., & Zeinal Hamadani, A. 2021, arXiv e-prints, arXiv:2111.03306

    2021 arXiv

  24. [32]

    & Clocchiatti, A

    Mortari, D. & Clocchiatti, A. 2007, Celestial Mechanics and Dynamical Astron- omy, 99, 45

    2007

  25. [33]

    Nawaz, A., Ahmad, A., & Khan, S. S. 2025, arXiv e-prints, arXiv:2509.07605

    2025

  26. [34]

    Offner, S. S. R., Moe, M., Kratter, K. M., et al. 2023, in Astronomical Society of the Pacific Conference Series, V ol. 534, Protostars and Planets VII, ed. S. Inutsuka, Y . Aikawa, T. Muto, K. Tomida, & M. Tamura, 275

    2023

  27. [35]

    2024, ApJ, 965, 89

    Pasquato, M., Trevisan, P., Askar, A., et al. 2024, ApJ, 965, 89

    2024

  28. [36]

    2011, Journal of Machine Learning Research, 12, 2825 Poincaré, H

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825 Poincaré, H. 1890, Acta Mathematica, 13, 1, oeuvres, tome VII, pp. 262-479 Portegies Zwart, S., Belleman, R., & Geldof, P. 2007, arXiv e-prints, cs/0702135 Portegies Zwart, ...

    2011 arXiv

  29. [37]

    & Liu, S.-F

    Rein, H. & Liu, S.-F. 2012, A&A, 537, A128

    2012

  30. [38]

    & Spiegel, D

    Rein, H. & Spiegel, D. S. 2015, MNRAS, 446, 1424

    2015

  31. [39]

    L., Weatherford, N

    Rodriguez, C. L., Weatherford, N. C., Coughlin, S. C., et al. 2022, ApJS, 258, 22

    2022

  32. [40]

    2018, ApJ, 853, 140 Santos Pereira, M., Tripa, L., Lima, N., Caldas, F., & Soares, C

    Samsing, J., MacLeod, M., & Ramirez-Ruiz, E. 2018, ApJ, 853, 140 Santos Pereira, M., Tripa, L., Lima, N., Caldas, F., & Soares, C. 2025, arXiv e-prints, arXiv:2503.04585

    2018 arXiv

  33. [41]

    & Kanisha, B

    Sasirekha, R. & Kanisha, B. 2025, Engineering Reports, 7, e70109

    2025

  34. [42]

    2025, ApJS, 281, 42

    SiTu, H., Wu, X.-J., Yang, B., & Lin, W. 2025, ApJS, 281, 42

    2025

  35. [43]

    H., & Bromm, V

    Stacy, A., Greif, T. H., & Bromm, V . 2010, MNRAS, 403, 45

    2010

  36. [44]

    2022, PASA, 39, e062

    Tory, M., Grishin, E., & Mandel, I. 2022, PASA, 39, e062

    2022

  37. [45]

    A., Spera, M., Leigh, N

    Trani, A. A., Spera, M., Leigh, N. W. C., & Fujii, M. S. 2019, ApJ, 885, 135

    2019

  38. [46]

    & Karttunen, H

    Valtonen, M. & Karttunen, H. 2006, The Three-Body Problem (Cambridge, UK: Cambridge University Press)

    2006

  39. [47]

    2018, arXiv e-prints, arXiv:1810.07168

    Wainer, J. 2018, arXiv e-prints, arXiv:1810.07168

    2018 arXiv

  40. [48]

    2020, MNRAS, 497, 536

    Wang, L., Iwasawa, M., Nitadori, K., & Makino, J. 2020, MNRAS, 497, 536

    2020

  41. [49]

    & Kroupa, P

    Weidner, C. & Kroupa, P. 2004, MNRAS, 348, 187

    2004

  42. [50]

    & Holman, M

    Wisdom, J. & Holman, M. 1991, AJ, 102, 1528

    1991

  43. [51]

    & Barba, L

    Yokota, R. & Barba, L. A. 2012, Computing in Science and Engineering, 14, 30 Article number, page 8 Ahmad Farhani Asl: Machine learning prediction of binary formation in three-body gravitational encounters Appendix A: Orbital Parameter Distributions Figure A.1 shows the distri...

    2012

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.